Chatbot teams everywhere want great virtual agents. But that assumes you know your customers’ goals for connecting with your chatbot. Getting it right can mean a good experience. Getting it wrong … that’s a different story.
But untangling intent performance is tough: figuring out where an intent goes wrong is manual, tedious and time-consuming. Conversely, evaluating an intent that performs according to the AI model can be a ruse – what if your bot performed as expected, but the customer decided to talk with a live agent anyway? Your analysis needs to include data points from that interaction as well.
So, the question is: how do you easily, and accurately, measure the performance of intent? And once you get this right, how do you monitor changes made in the system to confirm they are making a difference?
